intuition.

Understand it,
don't memorize it.

A community-built path through hard subjects, taught the way 3Blue1Brown teaches: visually, from the ground up, without skipping a single brick. Every step earns the next, and tells you exactly what to understand before it.

The journey

The Path to Machine Learning

Start with arrows on a grid. End understanding how a machine learns. Every step earns the next; no brick is skipped.

  1. 1
    Linear Algebra 22 concepts · start here

    Data is vectors and models are transformations. This is the ground everything else stands on.

  2. 2
    Calculus planned · open for a PR

    The math of change. A model learns by asking how a tiny nudge to its knobs changes its error.

    what a function is · the limit · the derivative · derivative as local zoom · the chain rule · partial derivatives · the gradient · integrals as accumulation

  3. 3
    Probability planned · open for a PR

    The math of uncertainty. Real data is noisy, and models reason in odds, not certainties.

    sample space and events · probability as measure · conditional probability · Bayes' theorem · random variables · distributions · expectation and variance · the normal distribution

  4. 4
    Optimization planned · open for a PR

    Learning is finding the lowest point of an error surface. This is how you get there.

    what optimization is · critical points · convexity · gradient descent · the learning rate · stochastic gradient descent · local vs global minima

    needs first: Calculus, Linear Algebra

  5. 5
    Statistics planned · open for a PR

    How to trust what data says, and how to catch a model that has fooled itself.

    population vs sample · estimation · bias and variance · sampling distributions · confidence intervals · hypothesis testing · overfitting

    needs first: Probability

  6. 6
    Machine Learning planned · open for a PR

    The destination. Every domain before it was already walking here.

    what learning means · linear regression · loss functions · training as optimization · classification · regularization · neural networks · backpropagation · PCA (eigenvectors, revisited)

    needs first: Linear Algebra, Calculus, Probability, Optimization, Statistics

Trust

No missing bricks

No concept assumes something you were never taught. Every prerequisite is a link, and it exists, checked automatically on every contribution. A roadmap with a missing brick is not a roadmap; it is a trap. The greyed steps above are the plan, waiting for someone to build them. Claim one.